Files
FileCodeBox/apps/base/file_validation.py
vastsa eda0a3127a fix: add magic-byte checks, login rate limits, and non-root Docker
Close the remaining security gaps: validate upload content against
file signatures, rate-limit admin login failures with configurable
thresholds, and run the container as uid 1000 instead of root.
2026-07-23 22:23:04 +08:00

197 lines
6.7 KiB
Python

"""文件类型校验:扩展名 / MIME 白名单 + magic bytes 防伪造。"""
from __future__ import annotations
from dataclasses import dataclass
from pathlib import Path
from typing import Optional
from fastapi import HTTPException, UploadFile
from fnmatch import fnmatchcase
from core.settings import settings
@dataclass(frozen=True)
class FileKind:
name: str
extensions: tuple[str, ...]
mimes: tuple[str, ...]
signatures: tuple[bytes, ...]
FILE_KINDS: tuple[FileKind, ...] = (
FileKind("png", (".png",), ("image/png",), (b"\x89PNG\r\n\x1a\n",)),
FileKind("jpg", (".jpg", ".jpeg"), ("image/jpeg",), (b"\xff\xd8\xff",)),
FileKind("gif", (".gif",), ("image/gif",), (b"GIF87a", b"GIF89a")),
FileKind("webp", (".webp",), ("image/webp",), ()),
FileKind("bmp", (".bmp",), ("image/bmp", "image/x-ms-bmp"), (b"BM",)),
FileKind("pdf", (".pdf",), ("application/pdf",), (b"%PDF",)),
FileKind(
"zip",
(".zip", ".docx", ".xlsx", ".pptx", ".apk", ".jar"),
(
"application/zip",
"application/x-zip-compressed",
"application/vnd.openxmlformats-officedocument.wordprocessingml.document",
"application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
"application/vnd.openxmlformats-officedocument.presentationml.presentation",
"application/java-archive",
"application/vnd.android.package-archive",
),
(b"PK\x03\x04", b"PK\x05\x06", b"PK\x07\x08"),
),
FileKind(
"rar",
(".rar",),
("application/x-rar-compressed", "application/vnd.rar"),
(b"Rar!\x1a\x07\x00", b"Rar!\x1a\x07\x01\x00"),
),
FileKind("7z", (".7z",), ("application/x-7z-compressed",), (b"7z\xbc\xaf'\x1c",)),
FileKind("gz", (".gz", ".tgz"), ("application/gzip", "application/x-gzip"), (b"\x1f\x8b",)),
FileKind(
"mp3",
(".mp3",),
("audio/mpeg",),
(b"ID3", b"\xff\xfb", b"\xff\xf3", b"\xff\xf2"),
),
FileKind("mp4", (".mp4", ".m4a", ".mov"), ("video/mp4", "audio/mp4", "video/quicktime"), ()),
FileKind(
"exe",
(".exe", ".dll", ".sys"),
("application/x-msdownload", "application/x-dosexec"),
(b"MZ",),
),
FileKind("elf", (".elf", ".so", ".o"), ("application/x-executable",), (b"\x7fELF",)),
)
KNOWN_EXTENSIONS = {ext for kind in FILE_KINDS for ext in kind.extensions}
def normalize_allowed_file_types() -> list[str]:
raw_value = settings.allowed_file_types
if isinstance(raw_value, str):
values = [item.strip() for item in raw_value.split(",")]
elif isinstance(raw_value, (list, tuple, set)):
values = [str(item).strip() for item in raw_value]
else:
values = []
normalized = [item.lower() for item in values if item]
return normalized or ["*"]
def _extension_of(file_name: str) -> str:
return Path(file_name or "").suffix.lower()
def _match_allow_rule(rule: str, file_name: str, content_type: str) -> bool:
if rule in {"*", "*/*"}:
return True
if "/" in rule:
return fnmatchcase(content_type, rule)
extension = rule if rule.startswith(".") else f".{rule}"
return file_name.endswith(extension)
def is_type_allowed(file_name: str, content_type: Optional[str] = None) -> bool:
allowed = normalize_allowed_file_types()
if any(rule in {"*", "*/*"} for rule in allowed):
return True
normalized_name = (file_name or "").strip().lower()
normalized_content_type = (content_type or "").strip().lower()
return any(
_match_allow_rule(rule, normalized_name, normalized_content_type)
for rule in allowed
)
def detect_file_kind(header: bytes) -> Optional[FileKind]:
if not header:
return None
if len(header) >= 12 and header.startswith(b"RIFF") and header[8:12] == b"WEBP":
return next(kind for kind in FILE_KINDS if kind.name == "webp")
if len(header) >= 12 and header[4:8] == b"ftyp":
return next(kind for kind in FILE_KINDS if kind.name == "mp4")
candidates: list[tuple[int, FileKind]] = []
for kind in FILE_KINDS:
for signature in kind.signatures:
if signature and header.startswith(signature):
candidates.append((len(signature), kind))
if not candidates:
return None
candidates.sort(key=lambda item: item[0], reverse=True)
return candidates[0][1]
def validate_file_type(file_name: str, content_type: Optional[str] = None) -> None:
if not is_type_allowed(file_name, content_type):
raise HTTPException(status_code=403, detail="不允许上传该类型文件")
def _kind_matches_ext(kind: FileKind, extension: str) -> bool:
return extension in kind.extensions
def _kind_matches_mime(kind: FileKind, mime: str) -> bool:
return mime in kind.mimes
def validate_file_magic(
file_name: str,
content_type: Optional[str],
header: Optional[bytes],
) -> None:
"""白名单通过后,用 magic bytes 防止扩展名 / MIME 伪造。"""
validate_file_type(file_name, content_type)
if not header:
return
claimed_ext = _extension_of(file_name)
claimed_mime = (content_type or "").strip().lower()
detected = detect_file_kind(header)
if claimed_ext in KNOWN_EXTENSIONS:
if detected is None or not _kind_matches_ext(detected, claimed_ext):
raise HTTPException(
status_code=403,
detail="文件内容与扩展名不匹配,疑似伪造类型",
)
if claimed_mime and any(claimed_mime in kind.mimes for kind in FILE_KINDS):
if detected is None or not _kind_matches_mime(detected, claimed_mime):
raise HTTPException(
status_code=403,
detail="文件内容与 Content-Type 不匹配,疑似伪造类型",
)
allowed = normalize_allowed_file_types()
if detected and not any(rule in {"*", "*/*"} for rule in allowed):
allowed_by_detected = any(
is_type_allowed(f"file{ext}", mime)
for ext in detected.extensions
for mime in (detected.mimes or (None,))
)
if not allowed_by_detected:
raise HTTPException(status_code=403, detail="不允许上传该类型文件")
async def read_upload_header(file: UploadFile, size: int = 64) -> bytes:
await file.seek(0)
header = await file.read(size)
await file.seek(0)
return header or b""
async def validate_upload_file(file: UploadFile) -> None:
header = await read_upload_header(file)
validate_file_magic(file.filename or "", file.content_type, header)
def validate_header_bytes(
file_name: str, content_type: Optional[str], header: bytes
) -> None:
validate_file_magic(file_name, content_type, header)